Google Professional Machine Learning Engineer Question 374
Select 3Google Cloud PlatformYou are a Machine Learning Engineer tasked with deploying a third-party MLFlow pipeline on Google Cloud. The solution must support model versioning, scalable serving, and integrations with Google Cloud's monitoring tools. Which combination of services should you use to achieve these requirements?
- A
Google Kubernetes Engine (GKE) to host the MLFlow server and ML pipelines
- B
Cloud Storage for storing MLFlow artifacts and model files
- C
Vertex AI Model Registry for model versioning and serving
- D
Cloud Monitoring for tracking the health and performance of the MLFlow deployment
- E
BigQuery ML for running the MLFlow pipeline directly
Show answer and explanation
Correct answers: A, B, D
Explanation
To deploy a third-party MLFlow pipeline on Google Cloud, you can use Google Kubernetes Engine (GKE) for hosting, Cloud Storage for artifact storage, and Cloud Monitoring for performance and health tracking. These services support scalability, integration with monitoring tools, and compatibility with MLFlow's architecture. While Vertex AI Model Registry provides model versioning, it is not suitable for directly managing MLFlow models, as MLFlow has its own versioning system.
- A. Correct.
Google Kubernetes Engine (GKE) can host custom MLFlow servers and pipelines, providing flexibility and scalability for third-party tools.
- B. Correct.
Cloud Storage is ideal for storing MLFlow artifacts such as models, datasets, and logs, ensuring compatibility with MLFlow's artifact storage requirements.
- C. Incorrect.
Vertex AI Model Registry is not compatible with MLFlow's native model versioning system. MLFlow has its own mechanism for managing model versions.
- D. Correct.
Cloud Monitoring integrates well with GKE and other Google Cloud services, enabling you to monitor the health and performance metrics of the MLFlow deployment.
- E. Incorrect.
BigQuery ML is designed for SQL-based machine learning workflows and cannot directly run third-party MLFlow pipelines.